This paper explores distributed aggregative games in multi-agent systems. Current methods for finding distributed Nash equilibrium require players to send original messages to their neighbors, leading to communication burden and privacy issues. To jointly address these issues, we propose an algorithm that uses stochastic compression to save communication resources and conceal information through random errors induced by compression. Our theoretical analysis shows that the algorithm guarantees convergence accuracy, even with aggressive compression errors used to protect privacy. We prove that the algorithm achieves differential privacy through a stochastic quantization scheme. Simulation results for energy consumption games support the effectiveness of our approach.
This paper considers a multi-user multiple-input multiple-output (MU-MIMO) system where the downlink communication between a base station (BS) and multiple user equipments (UEs) is aided by a reconfigurable intelligent surface (RIS). We study the sum rate maximization problem with the objective of finding the optimal precoding vectors and RIS configuration. Due to fronthaul limitation, each entry of the precoding vectors must be picked from a finite set of quantization labels. Furthermore, two scenarios for the RIS are investigated, one with continuous infinite-resolution reflection coefficients and another with discrete finite-resolution reflection coefficients. A novel framework is developed which, in contrast to the common literature that only offers sub-optimal solutions for optimization of discrete variables, is able to find the optimal solution to problems involving discrete constraints. Based on the classical weighted minimum mean square error (WMMSE), we transform the original problem into an equivalent weighted sum mean square error (MSE) minimization problem and solve it iteratively. We compute the optimal precoding vectors via an efficient algorithm inspired by sphere decoding (SD). For optimizing the discrete RIS configuration, two solutions based on the SD algorithm are developed: An optimal SD-based algorithm and a low-complexity heuristic method that can efficiently obtain RIS configuration without much loss in optimality. The effectiveness of the presented algorithms is corroborated via numerical simulations where it is shown that the proposed designs are remarkably superior to the commonly used benchmarks.
In this paper, we propose a novel semantic digital analog converter (sDAC) for the compatibility of semantic communications and digital communications. Most of the current semantic communication systems are based on the analog modulations, ignoring their incorporation with digital communication systems, which are more common in practice. In fact, quantization methods in traditional communication systems are not appropriate for use in the era of semantic communication as these methods do not consider the semantic information inside symbols. In this case, any bit flip caused by channel noise can lead to a great performance drop. To address this challenge, sDAC is proposed. It is a simple yet efficient and generative module used to realize digital and analog bi-directional conversion. On the transmitter side, continuous values from the encoder are converted to binary bits and then can be modulated by any existing methods. After transmitting through the noisy channel, these bits get demodulated by paired methods and converted back to continuous values for further semantic decoding. The whole progress does not depend on any specific semantic model, modulation methods, or channel conditions. In the experiment section, the performance of sDAC is tested across different semantic models, semantic tasks, modulation methods, channel conditions and quantization orders. Test results show that the proposed sDAC has great generative properties and channel robustness.
To satisfy the growing throughput demand of data-intensive applications, the performance of optical communication systems increased dramatically in recent years. With higher throughput, more advanced equalizers are crucial, to compensate for impairments caused by inter-symbol interference (ISI). The latest research shows that artificial neural network (ANN)-based equalizers are promising candidates to replace traditional algorithms for high-throughput communications. On the other hand, not only throughput but also flexibility is a main objective of beyond-5G and 6G communication systems. A platform that is able to satisfy the strict throughput and flexibility requirements of modern communication systems are field programmable gate arrays (FPGAs). Thus, in this work, we present a high-performance FPGA implementation of an ANN-based equalizer, which meets the throughput requirements of modern optical communication systems. Further, our architecture is highly flexible since it includes a variable degree of parallelism (DOP) and therefore can also be applied to low-cost or low-power applications which is demonstrated for a magnetic recording channel. The implementation is based on a cross-layer design approach featuring optimizations from the algorithm down to the hardware architecture, including a detailed quantization analysis. Moreover, we present a framework to reduce the latency of the ANN-based equalizer under given throughput constraints. As a result, the bit error ratio (BER) of our equalizer for the optical fiber channel is around four times lower than that of a conventional one, while the corresponding FPGA implementation achieves a throughput of more than 40 GBd, outperforming a high-performance graphics processing unit (GPU) by three orders of magnitude for a similar batch size.
The quality of wireless communication will directly affect the performance of federated learning (FL), so this paper analyze the influence of wireless communication on FL through symbol error rate (SER). In FL system, non-orthogonal multiple access (NOMA) can be used as the basic communication framework to reduce the communication congestion and interference caused by multiple users, which takes advantage of the superposition characteristics of wireless channels. The Minimum Mean Square Error (MMSE) based serial interference cancellation (SIC) technology is used to recover the gradient of each terminal node one by one at the receiving end. In this paper, the gradient parameters are quantized into multiple bits to retain more gradient information to the maximum extent and to improve the tolerance of transmission errors. On this basis, we designed the SER-based device selection mechanism (SER-DSM) to ensure that the learning performance is not affected by users with bad communication conditions, while accommodating as many users as possible to participate in the learning process, which is inclusive to a certain extent. The experiments show the influence of multi-bit quantization of gradient on FL and the necessity and superiority of the proposed SER-based device selection mechanism.
Despite the growing prevalence of large language model (LLM) architectures, a crucial concern persists regarding their energy and power consumption, which still lags far behind the remarkable energy efficiency of the human brain. Recent strides in spiking language models (LM) and transformer architectures aim to address this concern by harnessing the spiking activity of biological neurons to enhance energy/power efficiency. Doubling down on the principles of model quantization and energy efficiency, this paper proposes the development of a novel binary/ternary (1/1.58-bit) spiking LM architecture. Achieving scalability comparable to a deep spiking LM architecture is facilitated by an efficient knowledge distillation technique, wherein knowledge from a non-spiking full-precision "teacher" model is transferred to an extremely weight quantized spiking "student" LM. Our proposed model represents a significant advancement as the first-of-its-kind 1/1.58-bit spiking LM, and its performance is rigorously evaluated on multiple text classification tasks of the GLUE benchmark.
Segment Anything Model (SAM) has achieved impressive performance in many computer vision tasks. However, as a large-scale model, the immense memory and computation costs hinder its practical deployment. In this paper, we propose a post-training quantization (PTQ) framework for Segment Anything Model, namely PTQ4SAM. First, we investigate the inherent bottleneck of SAM quantization attributed to the bimodal distribution in post-Key-Linear activations. We analyze its characteristics from both per-tensor and per-channel perspectives, and propose a Bimodal Integration strategy, which utilizes a mathematically equivalent sign operation to transform the bimodal distribution into a relatively easy-quantized normal distribution offline. Second, SAM encompasses diverse attention mechanisms (i.e., self-attention and two-way cross-attention), resulting in substantial variations in the post-Softmax distributions. Therefore, we introduce an Adaptive Granularity Quantization for Softmax through searching the optimal power-of-two base, which is hardware-friendly. Extensive experimental results across various vision tasks (instance segmentation, semantic segmentation and object detection), datasets and model variants show the superiority of PTQ4SAM. For example, when quantizing SAM-L to 6-bit, we achieve lossless accuracy for instance segmentation, about 0.5\% drop with theoretical 3.9$\times$ acceleration. The code is available at \url{https://github.com/chengtao-lv/PTQ4SAM}.
The increasing size of large language models (LLMs) traditionally requires low-precision integer formats to meet strict latency and power demands. Yet recently, alternative formats such as Normal Float (NF4) have increased model accuracy at the cost of increased chip area. In this work, we first conduct a large-scale analysis of LLM weights and activations across 30 networks and conclude that most distributions follow a Student's t-distribution. We then derive a new theoretically optimal format, Student Float (SF4), that improves over NF4 across modern LLMs, for example increasing the average accuracy on LLaMA2-7B by 0.76% across tasks. Using this format as a high-accuracy reference, we then propose augmenting E2M1 with two variants of supernormal support for higher model accuracy. Finally, we explore the quality and efficiency frontier across 11 datatypes by evaluating their model accuracy and hardware complexity. We discover a Pareto curve composed of INT4, E2M1, and E2M1 with supernormal support, which offers a continuous tradeoff between model accuracy and chip area. For example, E2M1 with supernormal support increases the accuracy of Phi-2 by up to 2.19% with 1.22% area overhead, enabling more LLM-based applications to be run at four bits. The supporting code is hosted at https://github.com/cornell-zhang/llm-datatypes.
Scale is often attributed as one of the factors that cause an increase in the performance of LLMs, resulting in models with billion and trillion parameters. One of the limitations of such large models is the high computational requirements that limit their usage, deployment, and debugging in resource-constrained scenarios. Two commonly used alternatives to bypass these limitations are to use the smaller versions of LLMs (e.g. Llama 7B instead of Llama 70B) and lower the memory requirements by using quantization. While these approaches effectively address the limitation of resources, their impact on model performance needs thorough examination. In this study, we perform a comprehensive evaluation to investigate the effect of model scale and quantization on the performance. We experiment with two major families of open-source instruct models ranging from 7 billion to 70 billion parameters. Our extensive zero-shot experiments across various tasks including natural language understanding, reasoning, misinformation detection, and hallucination reveal that larger models generally outperform their smaller counterparts, suggesting that scale remains an important factor in enhancing performance. We found that larger models show exceptional resilience to precision reduction and can maintain high accuracy even at 4-bit quantization for numerous tasks and they serve as a better solution than using smaller models at high precision under similar memory requirements.
Text-driven human motion generation based on diffusion strategies establishes a reliable foundation for multimodal applications in human-computer interactions. However, existing advances face significant efficiency challenges due to the substantial computational overhead of iteratively solving for nonlinear reverse diffusion trajectories during the inference phase. To this end, we propose the motion latent consistency training framework (MLCT), which precomputes reverse diffusion trajectories from raw data in the training phase and enables few-step or single-step inference via self-consistency constraints in the inference phase. Specifically, a motion autoencoder with quantization constraints is first proposed for constructing concise and bounded solution distributions for motion diffusion processes. Subsequently, a classifier-free guidance format is constructed via an additional unconditional loss function to accomplish the precomputation of conditional diffusion trajectories in the training phase. Finally, a clustering guidance module based on the K-nearest-neighbor algorithm is developed for the chain-conduction optimization mechanism of self-consistency constraints, which provides additional references of solution distributions at a small query cost. By combining these enhancements, we achieve stable and consistency training in non-pixel modality and latent representation spaces. Benchmark experiments demonstrate that our method significantly outperforms traditional consistency distillation methods with reduced training cost and enhances the consistency model to perform comparably to state-of-the-art models with lower inference costs.
Large language models (LLMs) have revolutionized Natural Language Processing (NLP), but their size creates computational bottlenecks. We introduce a novel approach to create accurate, sparse foundational versions of performant LLMs that achieve full accuracy recovery for fine-tuning tasks at up to 70% sparsity. We achieve this for the LLaMA-2 7B model by combining the SparseGPT one-shot pruning method and sparse pretraining of those models on a subset of the SlimPajama dataset mixed with a Python subset of The Stack dataset. We exhibit training acceleration due to sparsity on Cerebras CS-3 chips that closely matches theoretical scaling. In addition, we establish inference acceleration of up to 3x on CPUs by utilizing Neural Magic's DeepSparse engine and 1.7x on GPUs through Neural Magic's nm-vllm engine. The above gains are realized via sparsity alone, thus enabling further gains through additional use of quantization. Specifically, we show a total speedup on CPUs for sparse-quantized LLaMA models of up to 8.6x. We demonstrate these results across diverse, challenging tasks, including chat, instruction following, code generation, arithmetic reasoning, and summarization to prove their generality. This work paves the way for rapidly creating smaller and faster LLMs without sacrificing accuracy.
Vector quantization, renowned for its unparalleled feature compression capabilities, has been a prominent topic in signal processing and machine learning research for several decades and remains widely utilized today. With the emergence of large models and generative AI, vector quantization has gained popularity in recommender systems, establishing itself as a preferred solution. This paper starts with a comprehensive review of vector quantization techniques. It then explores systematic taxonomies of vector quantization methods for recommender systems (VQ4Rec), examining their applications from multiple perspectives. Further, it provides a thorough introduction to research efforts in diverse recommendation scenarios, including efficiency-oriented approaches and quality-oriented approaches. Finally, the survey analyzes the remaining challenges and anticipates future trends in VQ4Rec, including the challenges associated with the training of vector quantization, the opportunities presented by large language models, and emerging trends in multimodal recommender systems. We hope this survey can pave the way for future researchers in the recommendation community and accelerate their exploration in this promising field.
Detection of changes in heterogeneous remote sensing images is vital, especially in response to emergencies like earthquakes and floods. Current homogenous transformation-based change detection (CD) methods often suffer from high computation and memory costs, which are not friendly to edge-computation devices like onboard CD devices at satellites. To address this issue, this paper proposes a new lightweight CD method for heterogeneous remote sensing images that employs the online all-integer pruning (OAIP) training strategy to efficiently fine-tune the CD network using the current test data. The proposed CD network consists of two visual geometry group (VGG) subnetworks as the backbone architecture. In the OAIP-based training process, all the weights, gradients, and intermediate data are quantized to integers to speed up training and reduce memory usage, where the per-layer block exponentiation scaling scheme is utilized to reduce the computation errors of network parameters caused by quantization. Second, an adaptive filter-level pruning method based on the L1-norm criterion is employed to further lighten the fine-tuning process of the CD network. Experimental results show that the proposed OAIP-based method attains similar detection performance (but with significantly reduced computation complexity and memory usage) in comparison with state-of-the-art CD methods.
The development of model compression is continuously motivated by the evolution of various neural network accelerators with ASIC or FPGA. On the algorithm side, the ultimate goal of quantization or pruning is accelerating the expensive DNN computations on low-power hardware. However, such a "design-and-deploy" workflow faces under-explored challenges in the current hardware-algorithm co-design community. First, although the state-of-the-art quantization algorithm can achieve low precision with negligible degradation of accuracy, the latest deep learning framework (e.g., PyTorch) can only support non-customizable 8-bit precision, data format, and parameter extraction. Secondly, the objective of quantization is to enable the computation with low-precision data. However, the current SoTA algorithm treats the quantized integer as an intermediate result, while the final output of the quantizer is the "discretized" floating-point values, ignoring the practical needs and adding additional workload to hardware designers for integer parameter extraction and layer fusion. Finally, the compression toolkits designed by the industry are constrained to their in-house product or a handful of algorithms. The limited degree of freedom in the current toolkit and the under-explored customization hinder the prototype ASIC or FPGA-based accelerator design. To resolve these challenges, we propose Torch2Chip, an open-sourced, fully customizable, and high-performance toolkit that supports user-designed compression followed by automatic model fusion and parameter extraction. Torch2Chip incorporates the hierarchical design workflow, and the user-customized compression algorithm will be directly packed into the deployment-ready format for prototype chip verification with either CNN or vision transformer (ViT). The code is available at https://github.com/SeoLabCornell/torch2chip.
We establish the fundamental limits in the approximation of Lipschitz functions by deep ReLU neural networks with finite-precision weights. Specifically, three regimes, namely under-, over-, and proper quantization, in terms of minimax approximation error behavior as a function of network weight precision, are identified. This is accomplished by deriving nonasymptotic tight lower and upper bounds on the minimax approximation error. Notably, in the proper-quantization regime, neural networks exhibit memory-optimality in the approximation of Lipschitz functions. Deep networks have an inherent advantage over shallow networks in achieving memory-optimality. We also develop the notion of depth-precision tradeoff, showing that networks with high-precision weights can be converted into functionally equivalent deeper networks with low-precision weights, while preserving memory-optimality. This idea is reminiscent of sigma-delta analog-to-digital conversion, where oversampling rate is traded for resolution in the quantization of signal samples. We improve upon the best-known ReLU network approximation results for Lipschitz functions and describe a refinement of the bit extraction technique which could be of independent general interest.
Fronthaul quantization causes a significant distortion in cell-free massive MIMO networks. Due to the limited capacity of fronthaul links, information exchange among access points (APs) must be quantized significantly. Furthermore, the complexity of the multiplication operation in the base-band processing unit increases with the number of bits of the operands. Thus, quantizing the APs' signal vector reduces the complexity of signal estimation in the base-band processing unit. Most recent works consider the direct quantization of the received signal vectors at each AP without any pre-processing. However, the signal vectors received at different APs are correlated mutually (inter-AP correlation) and also have correlated dimensions (intra-AP correlation). Hence, cooperative quantization of APs fronthaul can help to efficiently use the quantization bits at each AP and further reduce the distortion imposed on the quantized vector at the APs. This paper considers a daisy chain fronthaul and three different processing sequences at each AP. We show that 1) de-correlating the received signal vector at each AP from the corresponding vectors of the previous APs (inter-AP de-correlation) and 2) de-correlating the dimensions of the received signal vector at each AP (intra-AP de-correlation) before quantization helps to use the quantization bits at each AP more efficiently than directly quantizing the received signal vector without any pre-processing and consequently, improves the bit error rate (BER) and normalized mean square error (NMSE) of users signal estimation.
A fundamental problem associated with the task of network reconstruction from dynamical or behavioral data consists in determining the most appropriate model complexity in a manner that prevents overfitting, and produces an inferred network with a statistically justifiable number of edges. The status quo in this context is based on $L_{1}$ regularization combined with cross-validation. However, besides its high computational cost, this commonplace approach unnecessarily ties the promotion of sparsity with weight "shrinkage". This combination forces a trade-off between the bias introduced by shrinkage and the network sparsity, which often results in substantial overfitting even after cross-validation. In this work, we propose an alternative nonparametric regularization scheme based on hierarchical Bayesian inference and weight quantization, which does not rely on weight shrinkage to promote sparsity. Our approach follows the minimum description length (MDL) principle, and uncovers the weight distribution that allows for the most compression of the data, thus avoiding overfitting without requiring cross-validation. The latter property renders our approach substantially faster to employ, as it requires a single fit to the complete data. As a result, we have a principled and efficient inference scheme that can be used with a large variety of generative models, without requiring the number of edges to be known in advance. We also demonstrate that our scheme yields systematically increased accuracy in the reconstruction of both artificial and empirical networks. We highlight the use of our method with the reconstruction of interaction networks between microbial communities from large-scale abundance samples involving in the order of $10^{4}$ to $10^{5}$ species, and demonstrate how the inferred model can be used to predict the outcome of interventions in the system.
Recent transformer-based ASR models have achieved word-error rates (WER) below 4%, surpassing human annotator accuracy, yet they demand extensive server resources, contributing to significant carbon footprints. The traditional server-based architecture of ASR also presents privacy concerns, alongside reliability and latency issues due to network dependencies. In contrast, on-device (edge) ASR enhances privacy, boosts performance, and promotes sustainability by effectively balancing energy use and accuracy for specific applications. This study examines the effects of quantization, memory demands, and energy consumption on the performance of various ASR model inference on the NVIDIA Jetson Orin Nano. By analyzing WER and transcription speed across models using FP32, FP16, and INT8 quantization on clean and noisy datasets, we highlight the crucial trade-offs between accuracy, speeds, quantization, energy efficiency, and memory needs. We found that changing precision from fp32 to fp16 halves the energy consumption for audio transcription across different models, with minimal performance degradation. A larger model size and number of parameters neither guarantees better resilience to noise, nor predicts the energy consumption for a given transcription load. These, along with several other findings offer novel insights for optimizing ASR systems within energy- and memory-limited environments, crucial for the development of efficient on-device ASR solutions. The code and input data needed to reproduce the results in this article are open sourced are available on [https://github.com/zzadiues3338/ASR-energy-jetson].
Whisper is a multitask and multilingual speech model covering 99 languages. It yields commendable automatic speech recognition (ASR) results in a subset of its covered languages, but the model still underperforms on a non-negligible number of under-represented languages, a problem exacerbated in smaller model versions. In this work, we examine its limitations, demonstrating the presence of speaker-related (gender, age) and model-related (resourcefulness and model size) bias. Despite that, we show that only model-related bias are amplified by quantization, impacting more low-resource languages and smaller models. Searching for a better compression approach, we propose DistilWhisper, an approach that is able to bridge the performance gap in ASR for these languages while retaining the advantages of multitask and multilingual capabilities. Our approach involves two key strategies: lightweight modular ASR fine-tuning of whisper-small using language-specific experts, and knowledge distillation from whisper-large-v2. This dual approach allows us to effectively boost ASR performance while keeping the robustness inherited from the multitask and multilingual pre-training. Results demonstrate that our approach is more effective than standard fine-tuning or LoRA adapters, boosting performance in the targeted languages for both in- and out-of-domain test sets, while introducing only a negligible parameter overhead at inference.
Tiny machine learning (TinyML) for low-power devices lacks robust datasets for development. We present Wake Vision, a large-scale dataset for person detection that contains over 6 million quality-filtered images. We provide two variants: Wake Vision (Large) and Wake Vision (Quality), leveraging the large variant for pretraining and knowledge distillation, while the higher-quality labels drive final model performance. The manually labeled validation and test sets reduce error rates from 7.8% to 2.2% compared to previous standards. In addition, we introduce five detailed benchmark sets to evaluate model performance in real-world scenarios, including varying lighting, camera distances, and demographic characteristics. Training with Wake Vision improves accuracy by 1.93% over existing datasets, demonstrating the importance of dataset quality for low-capacity models and dataset size for high-capacity models. The dataset, benchmarks, code, and models are available under the CC-BY 4.0 license, maintained by the Edge AI Foundation.